Course talk:CPSC522/Recommendation System using Matrix Factorization

Hi Arthur, Great work on the page! I think your page is neat and very well structured. Some suggestions which I have are: 1. I really liked your gif on collaborative filtering. As we know, a picture or a demonstration says a thousand words; I would have enlarged it so the reader gets it in one go. Right now I had to click on it and was redirected to the file page and then seen the gif. You could manipulate the size using [[File:name.png|500px|thumb|right/center|Title] So the ‘500px’ will increase the size. You can modify accordingly.

2. I would do the same for the figure on 'Content based recommendation architecture’ and ’Two-level matrix multiplication'. So that the figure is readable along with the text.

3. Another thing I noticed is that I cannot access your references directly. I am first redirected to the paragraph on your page where you have referenced them and then clicking there I am able to go to that reference. I have used something like this to reference one of the wiki articles: {{Reflist|refs= *<ref name="ivsd">[https://en.wikipedia.org/wiki/Inductive_reasoning#Inductive_vs._deductive_reasoning Inductive vs Deductive reasoning]</ref>}} And i just use the ref name=“ivsd” to refer to this wherever I need to on my wiki page.

4. I would really like if you could add some applications or mention where collaborative filtering and content based filtering is being applied in real world examples, for instance Amazon etc. Hope this helps! Warm regards, Ritika

Dear Ritika

Thanks for your kindly detailed feedback. I have improve the wiki page of points 1,2 and 4 respectively by enlarging the page as well as the example of music genome project for the combination of collaborative filtering and content-based filtering. As for the reference part, I found error trying to use your example. I am still working on it.

Regards Arthur

Hi Arthur, Were you able to figure out the references part? Regards,

Ritika

Hi Sun,

very informative and easy to understand. And I am actually very interested in how could amazon always guess out exactly the book I am interested in. And now I know.

1. Can you introduce more about Textual Semantic Similarity and Weighted textual matrix factorization, although you introduced those, but still did not touch how it works.

2. I think you might need more info about two level Matrix Factorization, the first paper seems to have more information than this two level one

3. Do you want to make a conclusion to compare about the two different solutions?

Dear Dandan

Thank you for your great support and suggestions regarding my wiki pages.

I have already added the experimental result part at the end of the wiki page.

Regarding your suggestion 1 and 2, for suggestion 1, I think TSS and WTMF will account for a large portion of paragraph if elaborated in detail. In fact, the second paper just use it for the implicit information generation and it is not the most novelty point in the paper. I think the heuristic in this paper is to combine the implicit data into matrix factorization to solve the cold start problem and matrix factorization has already been detailed explained in the previous paper. So I didn't put too much emphasis on the second paper. Thank you

Regards Arthur

Hi Sun,

Nice work! Your page is informative and has nice examples. I have a few suggestions:

1. Highlight some key words for long paragraph then it is more easier to get the point as well as to read.
2. Some pictures in the page should be larger.
3. Add some explanation for pseudo code.
4. Put the formula to the centre of the line will make the page more beautiful.

Bests,

YuYan

Dear Yuanyu

Thank you very much for your great support and kind explanation. It really helps me a lot. Regarding for suggestions 1, 2 and 4. I have already modified that and changed accordingly. As for your suggestion 3, I previously thought about that but I didn't add that because I want to maintain the code brevity and the detail explanation of the code has already been explained in the previous theory part. Thank you.

Regards Arthur

Hi Sun,

Good work! Your page is very interesting and and informative.

My suggestions are:

• The images used in the page should be larger.
• Revise the page and correct the grammatical errors and spelling mistakes (eg. " Two-level matrix multipiliation " in the Implementation section figure.
• You could add a separate section at the end talking about how the second solution improved upon the first (i.e. describe the incremental contribution of one paper over the other)

Hope this helps,

Adnan

Dear Adnan

Thank you for your great support and kindly review as well as explanation. For your suggestion 1 and 2. I have already double checked to enlarge the picture and error spelling. As for the incremental contribution, please refer to the "What are the heuristic" section for the second paper contribution. Thank you

Regards Arthur